matich no.t pp. 125-134 4975
THE VALUE OF TOPOLOGICAL INDICES IN CHEMISTRY
D. H. Rouvray+ -Chemistry Department, University of the
Witwatersrand, Johannesburg, South Africa
In recent years it has become increasingly apparent
that many, if not all, physical and chemical properties
of fluids are dependent upon and often largely determined
by the topological structure of their component chemical
ew of this development, it would thus not
be unreaccnable to describe chemictry today ac an cocent-
species. In v
ially topological subject!. In this paper, ve briefly
review sexe of the principal areas of activity which have
led to this new situation. We start by tracing these
areas back to their origins
The earliest work along these lines, was that involving
studies of the additive properties cf compounds, initiated
during the first haif of the nineteenth century. As an
exemple, we cite the work cf Xopp?, who demonstrated that
the thermodynamic parameters of compounds such as boiling
point and specific heat were roughly additive in the eense
that an approxizately constant difference in the parameter
was observed for compounds differing by @ given chemical
grouping, such as a methylene unit. He was also the
et
person tc express his results in the form of algebraic
equations. During the succeeding decades, Kepr's researches
were extended by a variety cf vorkers, One of these,
Markownikey, showed that the elting point of a branched
hydrocarbon is higher then that for the corresponding
straight chein species?.- 126 ~
In the present century, Fiickel developed his now famous
procedures for obtaining appproximate answers to the
Schrédinger equation for hydrocarbon species". Extensions
to his work, by investigetors such as Coulson’, led to the
introduction of several new quantum chemical parameters
characteristic of hydrocarbon species, These included
the t-electronic charge density, the bond order, and the
polarizabilit;
y. In the early 1950's it was demonstrated
by Ruedenberg® that all such parameters were dependent
upon the molecular topology. Shortly thereafter, the
correspondence between the Hlckel matrix end what is now
referred to as the topological matrix was pointed out by
Alkanes to C,,- Viscosity at 308,15K
Standard Deviation = 0,052
log, (viscosity)
-05
fo&
log!
a
w
a oO
‘ogg!
~
- 127 -
Alkanes to C,,- Surface Tension at 29315K
Standard Deviation = 0,021
logg(surface tension)
28 29 30 31 32
Alkanes to C,- Refractive Index
Standard Deviation = 0,002
logglrefractive index)
29 30 31 32 33 3h- 128 -
Atkanes to C)- Boiling Point
6
5
2
8
42
3
2
Standard Deviation = 0,005
1
loge (boiling point)
l §2 54 56 53 60 62
Alkanes to C,.- Melting Point
6
a
4
3
2
Standard Deviation = 0,087
tog, (melting point)
48 49 §0 51 52-53 54 55 56- 129 -
Alkanes to Cj~ Density
6
log,
5
h
3
Standard deviation = 0,013
log, (density)
04% "05 06
Alkenes to C,,- Surface Tension at 293,15 K
7
6
5 ae
4
3
Standard Deviation = 0,050
log!
log, (surface tension)
25 26 27 28 29 30 “31 32 33- 130 -
Alkynes to C,,-Melting Point
loge!
a,
a
Standard Deviation = 0.064
loge (melting point)
56° 55 56
5) 52 53 57
loge!
o
Standard Deviation = 0,008
loge (boiling point)
58 59 60 6 62 63 64.65 66 67= 131 -
Glnthard and Prinas?. The Pairing Theorem is also realised
to be @ greph-theoretical result in essence®, Many original
contributions to this general area have been made within the
last five years by Gutman and Trinajstié?.
The evolution of topological indices in their own
right has come about during the last decade. The results
obtained in this area up to 1973 have been reviewed by the
present author!®, The fundamental idea in this approach is
that @ topological index of the structure of a component
species should reflect the bulk properties of the fluid
formed from the species in question. The condition that
all such species belong to a given chemical series, ensures
that interactions between species will ell be of the same
general type. By making this condition, intermolecular
forces may thereafter be neglected, at least to a first
approximation.
To date fairly good correlations have been obtained
with topological indices and the boiling points of members
of several homologous serics. Examples of euch correlations
are to be found in the works of Kosoya!! and Ranaiél?. Por
members of these series beyond about Cys, however, the
correlations have frequently been rather poor. In an
attempt to overcome these setbacks, the present author
developed a nev index, designed to be valid for a uch
wider range.
‘The new index proposed is based on the distance matrix
of a chemical species. The index is derived by simply
summing ell the entries in this matrix, and so the new
index is given by the summation
Although in such an index the structure within the matrix
is lost, it vas thought that this index might provide e
better representation of a species than some of the other
indices previously proposed, because it differentiates more
clearly betveen cinilar structures such as 2-nethyl hexane
and 3-nethyl hexane.
In practice, we have found good to excellent correl-- 132 -
ations between our index and the bulk parameters of iiquids.
In our vork we considered the members of the alkene, alkene,
and alkyne homologous series up to the Cyg members, and
arenes up to the Cag members. Using these molecules, ve
looked for correlations of our index against boiling point,
melting point, surface tension, viscosity, density, and
refractive index. A relationship of the general form
P= ky (x)#2
was assumed, where P is the bulk physicochemical parameter,
ky and kp are constants to be determined for a given series,
and I is our index, In all cases In P was plotted against
An I to test the reletionship.
Our investigations have revealed that in all cases «
correlation cf the type assumed does exist. Por the first
dozen or so merbers of all series investigated a fairly
good approximation to a straight line was obtained, The
best correlations were obtained for boiling points, and
the worst for melting points. Standard deviations vere
usually of the order cf 0,01 or less in this range. After
the first dozen or so members, the plots assumed a distinct
curvature, and standard deviations became correspondingly
poorer. All plots were found to be convex with respect to
the In I axis, vith the exception of the viscosity plots
which were always concave.
The shapes of the curves may be accounted for on the
following qualitative basis. We suppose that the molecular
topology is of overriding influence in determining physico-
chemical parameters vithin @ given series, as the inter-
molecular forces may ail be assumed of the same type. This
supposition explains in general terms the correlations ve
have found. But, because molecules are not rigid, we can
expect an increasing amount of folding and bending in the
liquid phase as the carbon content increases. As a result
of this, a relatively large molecule folded into the form
of a hairpin might behave as a molecule of only one half its
true length. Such folding would emount to an effective
loss of topological structure, which should become more
Pronounced vith increasing carbon content. This would lead- 133 -
to @ progressive lowering of the effective value of our
index, and-thus to the convex curves which we obtained.
In the case of the viscosity plots, a loss of topological
structure should result in @ reduced resistance to flow, and
therefore to an increased viscosity. On this basis a concave
curve is thus to be expected. Work is now undervay to make
these qualitative explanations quantitative by study of
simulated liquid structures, The relatively poor correl-
ations with melting points may be due to the fect that
lattice forces are superimposed upon the usual intermolecular
forces, and these to some extent modify the force field in
which the molecules find themselves. Lattice forces being
relatively independent of topological structure, it is only
to be expected thet poor correlations were found for this
parameter.
In conclusion, we may say that, because our correlations
were generally good, that the importance of topology as a
determinant of physicochemical behaviour has been established.
In fact, in some cases the correlations vere so good that it
would be feasible to make predictions of unknown parameters
from our index. Furthermore, the curvature found for all
our plots might be utilized to provide information on the
intermolecular forces in liquids and the relative arrangement
of the molecules in this phase. Up to the present, the only
type of parameters to have been investigated have been those
of a physical or chemical nature. In future, it may prove
possible to correlate biological parameters against e suitable
index. Work is at present underway in Johannesburg to
correlate the degree of carcinogenicity in arene species in
this way. It vould thus eppear that this fruitful new line
of inquiry is far from exhausted.- 134 -
References
lo.
ll.
R.W. Jotham, Chem. Soc. Reviews 2, 457 (1973).
H. Kopp, Ann, Chem. bi, 79, 169 (182).
V.V. Markovnikey, Ann. Chem. 162, 324 (1876).
EB. Hickel, Z. Phys. 70, 204, 279 (1931); idem. ibia.
12, 310 (1931).
C.A. Coulson, Proc. Roy. Soc. London A169, 413 (1939).
K. Ruedenberg, J. Chem. Phys. 22, 1878 (1954).
H.H. Gunthard and H. Primas, Helv. Chim. Acta 39,
1645 (1956).
D. Cvetkovié, Publ. Fac. Electrotechnique, Ser. Mat. Fiz.
(univ. Belgrade) 354-356, 1 (1971).
I. Gutman and N. Trinajstié, Topics in Curr. Chem. 42,
49 (1973).
D.H. Rouvray, Amer. Sci. 61, 729 (1973).
H. Hosoya, Bull. Chem. Soc. Japan Uh, 2332 (1971).
. Randié, J. Amer. Chem. Soc., to appear.